Five prominent publishers are filing a lawsuit against Meta regarding Llama.

      On Tuesday, Elsevier, Cengage, Hachette, Macmillan, and McGraw Hill, along with author Scott Turow, filed a proposed class action in Manhattan alleging that Meta illegally used millions of their works to train its Llama model. Following Judge Chhabria’s ruling in June 2025, plaintiffs with more substantial evidence of market harm have been awaiting their opportunity.

      That morning, five of the largest publishing companies in the world and a prominent American novelist entered a federal courthouse in Manhattan to file a proposed class action lawsuit against Meta Platforms. According to Reuters, the case involves the mentioned publishers and author Scott Turow claiming that Meta has unlawfully used millions of their books and journal articles to train its Llama large-language models without obtaining permission, payment, or a license. The lawsuit seeks court certification as a class action to represent all rights holders in similar situations.

      This lawsuit is the latest in a series of AI training copyright cases but differs significantly from most previous cases in substance.

      Why this case differs from Kadrey

      Those familiar with AI copyright litigation will recognize the precedent Kadrey v. Meta. Filed in 2023 in the Northern District of California, this case included authors like Sarah Silverman, Richard Kadrey, Christopher Golden, Ta-Nehisi Coates, Junot Díaz, and Michael Chabon, who made similar allegations against Meta for downloading copyrighted books from piracy sites like LibGen, Z-Library, and Anna’s Archive to train Llama.

      Court documents referenced by Tom’s Hardware revealed that Meta employees downloaded around 82 terabytes of pirated content during this process. Mark Zuckerberg reportedly approved the use of LibGen for Llama's training, despite internal concerns that it could jeopardize Meta’s negotiations with regulators.

      However, Meta won this case. In June 2025, Judge Vince Chhabria granted a summary judgment to Meta based on fair use, determining that the use of copyrighted material for Llama training met the fair-use criteria. Chhabria was notably specific about the limitations of his ruling, stating it might not align with reality and that it was only applicable to the authors involved in the case. He expressly noted that future plaintiffs could succeed with stronger market harm evidence, a critical aspect where he felt the Kadrey plaintiffs had faltered.

      The recent lawsuit appears to align with the kind of case Chhabria suggested could succeed.

      What the publishers offer that the authors lacked

      There are three key structural differences between the Kadrey case and this new lawsuit, all of which are advantageous for the plaintiffs. First is the catalog size. Kadrey involved about 666 specific titles from a limited set of authors, while the new complaint encompasses the entire publishing output of five companies that collectively represent a significant portion of global academic, educational, and trade publishing.

      According to Reuters, the titles in question include not only literary works like N.K. Jemisin's "The Fifth Season" and Peter Brown's "The Wild Robot," but also textbooks, scientific articles, and reference materials. The market dynamics for these works, especially in academic and educational sectors, differ from the trade fiction sector represented by the Kadrey plaintiffs.

      The second difference lies in the evidence of market harm. Academic and educational publishers can provide detailed documentation of revenue streams that AI-trained models effectively replace—unlike individual authors. For example, when Llama answers a student's biology question that would typically require consulting a Cengage textbook, that substitution is direct and quantifiable.

      The plaintiffs will likely argue that this substitution exemplifies the market harm that Chhabria indicated was lacking in the Kadrey case. An analysis by Reed Smith highlighted that market harm has emerged as a crucial element in fair-use decisions, more so than transformative use.

      Thirdly, the licensing-market context has evolved. Since 2023, AI companies have entered into a growing number of licensing agreements with publishers. Meta has established licensing arrangements with entities such as Reuters, CNN, Fox News, People Inc., and USA Today.

      The presence of these licenses is significant in fair-use law: courts assessing market harm now recognize that a licensing market exists, some publishers have negotiated terms within it, and Meta has opted to participate in certain markets while neglecting others. The new plaintiffs will argue that Meta’s selective licensing constitutes evidence of bad faith.

      The Anthropic settlement as a backdrop

      The Tuesday case occurs in light of another recent precedent. Earlier this year, Anthropic reached a settlement described by the Authors Guild as substantial, agreeing to compensate authors in the Bartz v. Anthropic class action concerning similar accusations. The settlement terms provide a benchmark for potential financial outcomes in AI training copyright cases.

      Meta faces a different scenario due to its unique circumstances and previous summary judgment win. The possibility of settlement exists, but Meta might also decide to defend the case similarly to Kadrey, relying on the fair-use argument even against more compelling plaintiffs. However, if the publishers were to lose a second time, it